What this actually is
Every serious AI lab now hands out a free tier. A few million tokens a month here, a few thousand requests a day there. On its own each one is a toy, which is why most developers sign up for one, hit a wall, and go pay for a plan.
FreeLLMAPI is an MIT licensed proxy that stacks them. You add the free keys you already qualify for, and it serves all of them behind a single OpenAI compatible endpoint. Point any OpenAI client at your local server and it routes transparently across whichever providers you have added.
- Repo: github.com/tashfeenahmed/freellmapi
- 18,265 stars, MIT licence, TypeScript
- 29 free providers, 251 model families, 358 free model endpoints
- Roughly 4 billion tokens per month of listed free capacity
- Keys are AES-256-GCM encrypted at rest and decrypted in memory per request
Two things to be clear about before you start
These are the two questions people ask within the first ten minutes, and getting them wrong wastes an evening.
- OpenAI is not one of the providers. GPT models are not part of the free pool. What you get is the free tiers of the labs and inference hosts that publish one.
- Claude appears here as a CLIENT, not as free inference. The router speaks Anthropic's wire format on /v1/messages, which means Claude Code and the official Anthropic SDKs can run against your free pool. It does not mean free Claude models.
The providers
These are the ones the README names directly. There are 19 more behind them, and the always current list with per model rate limits and context windows lives at freellmapi.co/models.
- Google, for the Gemini models
- Groq
- Cerebras
- OpenCode Zen
- Mistral
- OpenRouter
- Cloudflare
- Cohere
- Z.ai, for GLM
- NVIDIA
- HuggingFace
- ModelScope, for Qwen3, DeepSeek V4 and GLM-5, which needs an Aliyun CN binding
- Plus a custom provider, so you can point chat, embedding, image or audio models at any OpenAI compatible endpoint you already run, including llama.cpp, LM Studio, vLLM or a local Ollama
Free, for your email
Get the full setup steps + the copy-paste setup prompt
The rest of this guide (5 more sections) opens right here. The setup prompt, one paste into Claude that sets it all up, comes by email only. Use your real address.
Unlocks: Setting it up · Wiring up your agent · Why it does not fall over · The failure mode to plan for · A fair summary
50,000+ follow where these guides come from. No spam, reply “unsubscribe” any time.